Variable-size Memory Evolutionary Algorithm to Deal with Dynamic Environments: an empirical study



When dealing with dynamic environments two major aspects must be considered in order to improve the algorithms' adaptability to changes: diversity and memory. Integrating these two mechanisms in Evolutionary Algorithms we can enhance their performance. In this paper we propose and study a new EA that combines two populations, one playing the role of memory, with a biological inspired recombination operator to promote and maintain diversity. The size of the memory mechanism may vary along time. The size of the (usual) search population may also change in such a way that the sum of the individuals in the two populations does not exceed an established limit. The two populations have minimum and maximum lengths allowed and their sizes change according with the stage of the evolutionary process: if a alteration is detected in the environment, the search population increases its size in order to readapt quickly to the new conditions. When it is time to update memory, its size is increased if necessary. A genetic operator, inspired in the biological process of conjugation, is proposed and used combined with this memory scheme. Our ideas were tested under different dynamics and compared with other approaches on two benchmark problems. The obtained results show the efficacy, efficiency and robustness of the investigated algorithm.


Evolutionary algorithms, memory, diversity, dynamic environments


Evolutionary Optimization

TechReport Number

TR 2006/004

Cited by

No citations found